I played with the fragment mass tolerance but I am still having the issue.
However I searched the problematic files with TPP's X!Tandem (instead of
GPM's) and it worked perfectly with Peptide Prophet! Is there a linux
version of it which I can use on the cluster or is it just usable
Let me try lowering fragment mass error. I will let you know about the
Well, we already had the Proteome Discoverer results and we really want to
create the spectral library asap, so I was given those results to use
instead of X!Tandem. and I have never worked with
The TPP is not tested to work with ProteomeDiscoverer pepXML. Sure you can
send me a sample file, but it may take some time to implement support for
this type of data. Comet is open source and easy to use out of the box.
Why don't you try that route?
On Wed, Oct 12, 2016 at 3:24 PM, Ali
Thank you very much! You were right, the problem was with X!Tandem. I
decided do the searches with Proteome Discoverer. I exported the results to
pep.xml (from .msf) using PD but now the results don't work with
PeptideProphet. I can open them in PepXMLViewer but as I said in another
My suspicion regarding the Tandem failure is the fragment mass tolerance.
I am observing fragment mass differences of much less than 0.05 daltons,
yet your Tandem search uses 0.5 daltons as the tolerance. Could GPM-tandem
be having a problem there? Can you try to lower the fragment mass
I searched your data with Comet and PeptideProphet got valid PSMs in the
analysis that failed to produce correct results from your GPM-version of
tandem on the same data. The problem in your dataset is the Tandem
search. Attached are the comet parameters I used to search your data.
I sent the link to your email. Thank you very much!
On Wednesday, October 5, 2016 at 9:21:04 PM UTC-4, David Shteynberg wrote:
> I cannot tell from the search parameters if there is a problem. The only
> issue I saw was that you didn't enable isotopic offsets, which are
> sometimes helpful.
How can I send you these files? Can I upload them on the cloud and send you
I am not doing the mass spec so I don't know the details of each
experiments but some of them have all three charges and some of them only
have 2+ and 3+. Here is an example of similar issue for a file with
I cannot tell from the search parameters if there is a problem. The only
issue I saw was that you didn't enable isotopic offsets, which are
sometimes helpful. If you send me a sample mzML file and possibly the
search results in pep.xml along with the database I can troubleshoot
We are only collecting 2+, 3+ and 4+ data. I am working on the cluster, so
I am using the GPM version and we were unable to install the k-score plugin
therefore I am using the X! Tandem native scoring for the searches.
If you mean combining all of the search results with xinteract
Are you setting your instrument to collect only 2+ and 3+ data? Are you
using the tandem bundled with TPP or the GPM version of tandem? Have you
tried to process all your runs in one file?
On Wed, Oct 5, 2016 at 11:39 AM, Ali wrote:
Thank you very much! I really appreciate it!
On Tuesday, October 4, 2016 at 8:03:27 PM UTC-4, David Shteynberg wrote:
> PeptideProphet is doing it's job here by telling you that there is no
> distinct population of correct results in this search. I suspect your
> search parameters are the
PeptideProphet is doing it's job here by telling you that there is no
distinct population of correct results in this search. I suspect your
search parameters are the problem. This is orbitrap data so the charges
should be known, but it seems Tandem only searched 2+ and 3+ for you, this
It appears that PeptideProphet fails, I am getting this message:
using Accurate Mass Bins
using PPM mass difference
(X! Tandem) (minprob 0)
adding Accurate Mass mixture distr
init with X! Tandem trypsin
MS Instrument info: Manufacturer: Thermo Scientific, Model: UNKNOWN,
Thanks for your recommendations David.
I added decoys to my data base and I am still getting 0 probabilities for
all of the hits. I am attaching my X! Tandem search parameters xml, could
you please kindly do me a big favor and check it to see if there is
anything suspicious there which might
I highly recommend you start employing decoys in your search databases. I
noticed you are not combining together the runs in the analysis. If the
data was generated and searched in the same way it should be analysed
together for better statistical power. I suggest you combine the results
Is it possible there are no correct results in that set? Common pitfalls
are incorrect search parameters (e.g. wrong mods enabled or disabled) and
wrong database. If you have decoys in your database these can be utilized
for a more accurate mixture model generation using options "DECOY=
Mail list logo